Coding Challenge - Deep Learning for NLP (Foong)

Description:

This repository contains a Jupyter notebook using scikit-learn SVM to classify real & fake news.

Dataset: https://www.kaggle.com/clmentbisaillon/fake-and-real-news-dataset Libraries used: Scikit-learn, NLTK, pandas, numpy, csv

Write-up:

The accuracy of the model is 0.995.

There are a couple of misclassified news articles and to improve the model's performance on these news articles, here're some suggestions:

• Remove stop words: The news article title and text contain a lot of commonly used words which should be removed as features. Therefore, more data cleaning should be performed prior to model building.
• Try using the neural network by setting batch size, apply dropout & finetuning it
• Run cross-validation